[Battlemesh] BMV8 Final blog post

Dave Taht dave.taht at gmail.com
Fri Aug 14 15:42:37 CEST 2015

well, there was a third testbed, but it was used in production to
carry traffic from the lecture hall to the rest of the internet -
utilizing the new minstrel blues and minstrel variance patches. (this
was the "babel" ssid)

I did not get a lot of data from it during the conference, my goal was
merely to determine if minstrel-blues co-existed with other
non-power-reducing stations in a busy wifi environment. The potential
problem here was similar to using delay based tcps vs loss based -
loss based always wins.

it worked pretty darn well during the conference - but more exaustive
analysis is needed.

Prior to the conference I did get a lot of data from it, that is
difficult to interpret. It may well be that that andrew's
minstrel-variance patches result in too much stability at a given rate
and it needs to probe harder to find better rates than it currently
does. I see *very good* performance but rate switches only seem to
take place on roughly 30 second intervals.

the topology of that testbed was:

wlan-slovenia -ethernet- wndr4300 - nanostation m5 - nanostation m5 -
"babel-lecture" ssid on a picostation.

with a few other picostations and nanostations thrown in for good measure.

On Fri, Aug 14, 2015 at 6:27 AM, Juliusz Chroboczek
<jch at pps.univ-paris-diderot.fr> wrote:
>>> This is a story of at least 2 testbeds,
>> No, same testbed. All credits for setting that up to the manual testbed
>> team; I just ran a couple of tests on it :)
> To clarify, just in case anyone is still confused:
>   - the automated WiBed testbed never managed to produce any data, which
>     is a tragic waste of the time and work of a number of good people;
>   - the manual testbed produced two sets of results:
>       - the first one using some ad-hoc scripts;
>       - the second one using Toke's "flent" tool.
> We will work on making a nice web page with the results when Toke is in
> Paris next week.
> I am a strong believer in giving proper credit where credit is due, so
> please let me repeat that I am aware of the following people having worked
> on the manual testbed (in alphabetical order):
>   Amadeus, Dario, Federico, Goran, Henning, Juliusz, Matthieu, Riccardo,
>   Thijs, Tomislava, Toke.
> _______________________________________________
> Battlemesh mailing list
> Battlemesh at ml.ninux.org
> http://ml.ninux.org/mailman/listinfo/battlemesh

Dave Täht
worldwide bufferbloat report:
What will it take to vastly improve wifi for everyone?

More information about the Battlemesh mailing list